Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TravelAgent: An AI Assistant for Personalized Travel Planning

Aili Chen, Xuyang Ge, Ziquan Fu, Yanghua Xiao, Jiangjie Chen

TL;DR

TravelAgent addresses the challenge of generating realistic, comprehensive, and personalized travel itineraries under multi-dimensional constraints. It combines four modules—Tool-usage, Recommendation, Planning, and Memory—into a spatiotemporal framework where constraints are modeled, real-time data is integrated via tools, and user preferences are learned over time to tailor recommendations and route plans. Key contributions include a constraint-aware planning algorithm with a Budget Planner and Route Planner, an online recommendation framework leveraging in-context constraints, real-time information, and memory-driven insights, and extensive evaluation showing improvements in Rationality, Comprehensiveness, and Personalization over baselines. The results demonstrate practical impact for automated travel planning, enabling dynamic, user-aware itineraries in real-world settings, while highlighting the need for reliable data sources and deeper personalization in future work.

Abstract

As global tourism expands and artificial intelligence technology advances, intelligent travel planning services have emerged as a significant research focus. Within dynamic real-world travel scenarios with multi-dimensional constraints, services that support users in automatically creating practical and customized travel itineraries must address three key objectives: Rationality, Comprehensiveness, and Personalization. However, existing systems with rule-based combinations or LLM-based planning methods struggle to fully satisfy these criteria. To overcome the challenges, we introduce TravelAgent, a travel planning system powered by large language models (LLMs) designed to provide reasonable, comprehensive, and personalized travel itineraries grounded in dynamic scenarios. TravelAgent comprises four modules: Tool-usage, Recommendation, Planning, and Memory Module. We evaluate TravelAgent's performance with human and simulated users, demonstrating its overall effectiveness in three criteria and confirming the accuracy of personalized recommendations.

TravelAgent: An AI Assistant for Personalized Travel Planning

TL;DR

TravelAgent addresses the challenge of generating realistic, comprehensive, and personalized travel itineraries under multi-dimensional constraints. It combines four modules—Tool-usage, Recommendation, Planning, and Memory—into a spatiotemporal framework where constraints are modeled, real-time data is integrated via tools, and user preferences are learned over time to tailor recommendations and route plans. Key contributions include a constraint-aware planning algorithm with a Budget Planner and Route Planner, an online recommendation framework leveraging in-context constraints, real-time information, and memory-driven insights, and extensive evaluation showing improvements in Rationality, Comprehensiveness, and Personalization over baselines. The results demonstrate practical impact for automated travel planning, enabling dynamic, user-aware itineraries in real-world settings, while highlighting the need for reliable data sources and deeper personalization in future work.

Abstract

As global tourism expands and artificial intelligence technology advances, intelligent travel planning services have emerged as a significant research focus. Within dynamic real-world travel scenarios with multi-dimensional constraints, services that support users in automatically creating practical and customized travel itineraries must address three key objectives: Rationality, Comprehensiveness, and Personalization. However, existing systems with rule-based combinations or LLM-based planning methods struggle to fully satisfy these criteria. To overcome the challenges, we introduce TravelAgent, a travel planning system powered by large language models (LLMs) designed to provide reasonable, comprehensive, and personalized travel itineraries grounded in dynamic scenarios. TravelAgent comprises four modules: Tool-usage, Recommendation, Planning, and Memory Module. We evaluate TravelAgent's performance with human and simulated users, demonstrating its overall effectiveness in three criteria and confirming the accuracy of personalized recommendations.
Paper Structure (24 sections, 5 figures, 3 tables, 1 algorithm)

This paper contains 24 sections, 5 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. LLM-based Agent
  4. LLM as Recommendation System
  5. The TravelAgent System Design
  6. Constraints Modeling
  7. Tool-usage Module
  8. Recommendation Module
  9. Planning Module
  10. Budget Planner
  11. Route Planner
  12. Attraction Scores Calculation
  13. Memory Module
  14. Evaluation
  15. Overall Evaluation
  16. ...and 9 more sections

Figures (5)

  • Figure 1: Initial constraints modeling with user input and Memory Module.
  • Figure 2: An example of Hotel Tool call. Appendix \ref{['appendix:tools']} provides a comprehensive overview of all types and functions of tools.
  • Figure 3: The process of Attraction Recommendation.
  • Figure 4: The whole workflow of Route Planner illustrates the dynamic generation process of multi-day itineraries.
  • Figure 5: MAE and RMSE trends over simulated user interactions. When with user behavior history, our methods demonstrate superior cold-start performance and excellent error control. When without user behavior history, our methods mitigate performance degradation effectively compared to baselines. Prompts are shown in Appendix \ref{['appendix:prompts']}.